Improvement of Thermoelectric Performance of CoSb3−xTex

State Key Laboratory of New Ceramics and Fine Processing, Department of Materials Science and Engineering, Tsinghua University, Beijing 100084, China,...
0 downloads 0 Views 1MB Size
7526

Chem. Mater. 2008, 20, 7526–7531

Improvement of Thermoelectric Performance of CoSb3-xTex Skutterudite Compounds by Additional Substitution of IVB-Group Elements for Sb Wei-Shu Liu,†,‡ Bo-Ping Zhang,‡ Li-Dong Zhao,†,‡ and Jing-Feng Li*,† State Key Laboratory of New Ceramics and Fine Processing, Department of Materials Science and Engineering, Tsinghua UniVersity, Beijing 100084, China, and School of Materials Science and Engineering, UniVersity of Science and Technology Beijing, Beijing 100083, China ReceiVed September 2, 2008. ReVised Manuscript ReceiVed October 28, 2008

An enhanced ZT value as high as 1.1 at ∼550 °C has been achieved in n-type nanostructured CoSb3-xTex skutterudite compounds through additional substitution of IVB-group elements (Si, Ge, Sn, Pb) for Sb. Particularly, Sn was found to be the most effective element to enhance the ZT value by extending the solubility limit of Te in CoSb3-xTex, which decreases thermal conductivity more significantly than electrical conductivity. The reduced thermal conductivity is confirmed by the Raman scattering measurement of CoSb2.86M0.02Te0.12 (M ) Si, Ge, Sn, Pb), which shows that Ge and Sn enter into the Sb-site of CoSb3 and generate significant changes in the vibration modes, whereas Si and Pb are not likely to get into the lattice of CoSb3. Furthermore, nanostructure with fine grains and “nanodots” also contributes to the reduction of thermal conductivity.

Introduction Thermoelectric energy conversion technique has received interests because of their promising application in directly converting both the existing heat in nature and waste heat generated in industry. However, the low efficiency limits their applications to some cases where reliability is more important than efficiency, such as space missions.1 The efficiency of thermoelectric energy conversion is significantly controlled by the dimensionless figure of merit ZT ) T(σR2)/λ of thermoelectric materials, where T is the absolute temperature, R the Seebeck coefficient, σ the electrical conductivity, and λ the thermal conductivity that is a sum of carrier component λcar and lattice component λlat. However, the interdependence of these three physical parameters complicates the task of enhancing ZT value. An increase in σ usually leads to a decrease in R and an increase in λcar. An introduction of phonon scattering mechanism to reduce λlat often simultaneously produces scattering on the electrons and hence decreases σ. Therefore, the challenge for the development of high-performance thermoelectric materials is to reduce λlat significantly while avoiding obvious deterioration of σ.2,3 Among many thermoelectric materials, CoSb3-based skutterudites have received considerable interests because of their unique crystal structure and good thermoelectric properties.4 * Corresponding author. E-mail: [email protected]. † Tsinghua University. ‡ University of Science and Technology Beijing.

(1) Yang, J.; Caillat, T. MRS Bull. 2006, 31, 224. (2) Dresselhaus, M. S.; Chen, G.; Tang, M. Y.; Yang, R. G.; Lee, H.; Wang, D. Z.; Ren, Z. F.; Fleurial, J. P.; Gogna, P. AdV. Mater. 2007, 19, 1043. (3) Snyder, G. J.; Toberer, E. S. Nat. Mater. 2008, 7, 105. (4) Uher, C. Skutterudite: ProspectiVe NoVel Thermoelectrics in Recent Trends in Thermoelectric Materials Research I; Tritt, T. M, Ed.; Semiconductors and Semimetals; Academic Press: San Diego, 2001; Vol. 69, p139.

There are two chemical approaches to enhance the thermoelectric performance of CoSb3-based alloys: one is filling the lattice cage of skutterudite structure, and another is elemental substitution. Many studies so far have been conducted on filled CoSb3 compounds, 5,6 which are limited to some specific filling elements and melting processing. As compared with filled CoSb3-based alloys, there are more elements which can be used to substitute Co and/or Sb elements,7,8 and most processes can be applied to its synthesis.9-12 In our previous study,13 the effect of Te substitution in CoSb3 was confirmed, and a high ZT∼0.9 was achieved in nanostructured CoSb3-xTex fabricated by a powder metallurgy process combining mechanical alloying (MA) and spark plasma sintering (SPS). However, the solubility of Te in CoSb3-xTex is quite low (∼5%) because of the charge difference between Te and Sb. This charge difference can be compensated by doping elements with less valence electron than host element Sb. In this study, IVB-group elements (Si, Ge, Sn, Pb) were intentionally chosen as charge compensating element for Te (5) Chen, L. D.; Kawahara, T.; Tang, X. F.; Goto, T.; Hirai, T.; Dyck, J. S. W.; Chen, W.; Uher, C. J. Appl. Phys. 2001, 90, 1864. (6) Nolas, G. S.; Kaeser, M.; Littleton IV, R. T.; Tritt, T. M. Appl. Phys. Lett. 2000, 77, 1855. (7) Caillat, T.; Borshchevsky, A.; Fleurial, J. P. J. Appl. Phys. 1996, 80, 4442. (8) Liu, W. S.; Zhang, B. P.; Li, J. F.; Zhang, H. L.; Zhao, L. D. Acta Phys. Sin. 2008, 57, 3791. (9) Tang, X. F.; Chen, L. D.; Goto, T.; Hirai, T.; Yuan, R. Z. Acta Phys. Sin. 2000, 49, 1120. (10) Toprak, M. S.; Stiewe, C.; Platzek, D.; Williams, S.; Bertini, L.; Mu¨ller, E.; Gatti, C.; Zhang, Y.; Rowe, M.; Muhammed, M. AdV. Funct. Mater. 2004, 12, 1189. (11) Liu, W. S.; Zhang, B. P.; Li, J. F.; Zhao, L. D. J. Phys. D: Appl. Phys. 2007, 40, 6784. (12) Liu, W. S.; Zhang, B. P.; Li, J. F.; Zhao, L. D. J. Phys. D: Appl. Phys. 2007, 40, 566. (13) Liu, W. S.; Zhang, B. P.; Li, J. F.; Zhang, H. L.; Zhao, L. D. J. Appl. Phys. 2007, 102, 103717.

10.1021/cm802367f CCC: $40.75  2008 American Chemical Society Published on Web 12/03/2008

Thermoelectric Performance of CoSb3-xTex

Chem. Mater., Vol. 20, No. 24, 2008 7527

to raise the solubility of Te in CoSb3-xTex. We found that a high ZT of 1.1 at 550 °C was achieved in the Sn/Te-codoped CoSb3 because of the significant decrease in lattice thermal conductivity but slight decrease in carrier mobility. The mechanism for improving the thermoelectric performance of CoSb3 alloys by mixed elemental substitution for the Sb site has been discussed on the basis of the Raman scattering measurements and nanostructure observation experiments. Experimental section

Figure 1. XRD patterns of CoSb2.75Sn0.25-xTex (x ) 0.125 0.150, 0.175, 0.200). The standard data of CoSb2 and CoTe2 are referenced from the JCPDS Card Nos. PDF 65-4102 and PDF 89-2091, respectively.

Synthesis. In the CoSb3-y-xMyTex family, both n-type and p-type samples can be obtained by tuning the parameters x and y. In this study, two series of n-type CoSb3-xMyTex (one is M ) Sn, y ) 0.25 - x, x ) 0.125, 0.150, 0.175, 0.200; other is M ) Si, Ge, Sn, Pb, y ) 0.02, x ) 0.12) samples were focused. The fabrication process used was similar to that reported in our previous studies.8,13 A stoichiometric powder mixtures of Co, Sb, Te, and M (including Si, Ge, Sn, Pb) were subjected to mechanical alloying for 15 h at 450 rpm, and then sintered at 500 °C for 5 min by using a SPS system (Sumitomo SPS1050, Japan) under 50 MPa. X-ray Diffraction Pattern. The phase constitutions were characterized by X-ray diffraction instrument (XRD, Rigaku2500, Cu KR, Japan), which was operated at 40kV and 20mA. Lattice constant of CoSb3 phase was determined from the data in the range of 2θ between 20° and 100° by an analytical extrapolation function of (cos2 θ/sin θ + cos2 θ/θ)/2. Field-Emission Scanning Electron Microscopy. The microstructure and chemical homogeneity of the samples were examined by field emission scanning electron microscopy (FE-SEM, JSM6301F, Japan). Specimens used for the investigation are fixed on the copper holder by silver paste to avoid the specimen drift during the energy dispersive X-ray (EDX) measurement process. High-Resolution Transmission Electron Microscopy. The nanostructure of some samples was examined using transmission electron microscopy (TEM, JSM-2011, Japan). Specimens used for the investigation were polished to about 50-60 µm with sand paper and subsequently diamond paste. Specimens were then thinned to electron transparency using a Gatan model 600CDIF precession ion-milling system at low angle (10-12°). HRTEM images of several different locations of the specimens were obtained at 200 kV. Thermoelectric Transport Property. The Seebeck coefficient and electrical resistivity were measured using a Seebeck Coefficient/ Electrical Resistivity Measuring System (ZEM-2, Ulvac-Riko, Japan). The thermal conductivity was calculated by the relationship of λ ) DCpd, where D, Cp, and d are the thermal diffusivity, heat capacity, and density, respectively. The thermal diffusivity D was measured by the laser flash method (Netzsch Laser Flash Apparatus LFA 427, Germany), and the density was measured by Archimedes method. The heat capacity used in this study is calculated data of pure CoSb3,14 which is given in the Supporting Information. The room-temperature heat capacity of doped CoSb3 compounds were also confirmed using a differential scanning calorimeter (Dupont 1090B, USA). The experimental heat capacities are consistent with the calculated data, showing a little contribution by the composition change. Raman Scattering Spectroscopy. To identify the effect of the group IV elements on the vibration modes, we investigated the Raman scattering response of CoSb2.86M0.02Te0.12 (M ) Si, Ge, Sn,

Phase and Microstructure. Figure 1 shows the XRD patterns of CoSb2.75Sn0.25-xTex (x ) 0.125, 0.150, 0.175, 0.200). The XRD patterns demonstrate that the major phase is CoSb3 phase, which is crystallized in a cubic CoAs3-type structure with IM-3 space group. A small amount of CoSb2 and CoTe2 were also observed, which is similar to the previous case in CoSb3-xTex,13 but no diffraction peak of SnTe was detected. The XRD peaks of impurity slightly deviated from standard XRD patterns of CoSb2 (PDF No. 65-4102) and CoTe2 (PDF No. 89-2091), which is caused by the fact some Te/Sn atoms entered into the impurity phases. Takizawa et al. 15 reported that group IV elements, including Si, Ge, Sn and Pb, can enter into the cagelike 2a site of CoSb3 lattice under a ultrahigh pressure of several GPa, which can be judged by the decreased ratio of (211)/ (310) of XRD patterns. However, no such changes were observed in the present samples. In contrast, a p-type semiconductor was obtained as the concentration of Sn was much more than that of Te. The group IVB elements are likely to enter into the Sb-occupying 8c-site rather than the cagelike 2a site. The lattice constant increased almost linearly from ∼9.0350 Å for pure CoSb3 to 9.0413, 9.0428, 9.0433, and 9.0452 Å, respectively for x ) 0.125, 0.150, 0.175, and 0.200 compositions, suggesting that additional Te atoms exceeding the solubility limit (x ) ∼0.150) in CoSb3-xTex successfully entered into the Sb-site of CoSb3 lattice due to the coaddition of Sn. It is known that the doping solubility is affected by both the size difference and charge difference. The valence electron number (atom size) is 4 (1.72 Å), 5 (1.53 Å), 6 (1.42 Å), respectively, for Sn, Sb, Te. The addition of Sn not only compensates for the charge difference, but also relaxes the lattice stress due to the size difference between Te and Sb in CoSb3-xTex. We also noted that the lattice constant of CoSb3-xTex was slightly decreased

(14) Barin, I. Thermochemical Data of Pure Substrate (in Chinese); China Science Press: Beijing, 2003.

(15) Takizawa, H.; Miura, K.; Ito, M.; Suzuki, T.; Endo, T. J. Alloys Compd. 1999, 282, 79.

Pb) samples. The room-temperature Raman scattering measurements were performed by using a microconfocal Raman spectrometer (RM2000, Reishaw, England). The laser excitation used was 514.5 nm line of an Ar+-ion laser with a typical laser spot size of 2 µm. The Raman scattering data was collected by using a charge coupled device detector array with a typical resolution of 1.7 wave numbers.

Results and Discussion

7528

Chem. Mater., Vol. 20, No. 24, 2008

Liu et al.

Figure 2. Microstructure and elemental distribution of the CoSb2.75Sn0.05Te0.20 sample: (a) TEM image, (b) FE-SEM image, (c) EDX detecting region, (d) distribution of element Sn, and (e) distribution of element Te.

due to the additional substitution by Sn. For example, the lattice constant for CoSb2.85Te0.15 is 9.0447 Å,13 whereas it is 9.0428 Å for CoSb2.75Sn0.01Te0.15. At present, it is not well understood why the lattice constant of CoSb3 increased when doped with smaller Te than Sb but decreased when doped with larger Sn, but such an abnormal change may be related to the special lattice structure of CoSb3. Figure 2 shows a representative microstructure and elemental distribution in submicro scale for the CoSb2.75Sn0.05Te0.2 sample. We can see well-developed angular grains and clear grain boundaries from the TEM image (Figure 2a) and FE-SEM image (Figure 2b). The present average grain size is about 140 nm, which is slightly smaller than that (160 nm) of CoSb2.85Te0.15.13 Such a finegrained microstructure, which is in favor of thermoelectric materials requiring low thermal conductivity, was obtained because MA can produce fine particles, whose growth is subsequently suppressed during the SPS consolidation process. In order to investigate the chemical homogeneity of Sn/Te doped CoSb3, we mapped the elemental distributions of CoSb2.75Sn0.05Te0.2 by EDX, within an area shown in Figure 2c, and it was confirmed that the elemental distributions of Co, Sn, Sb, Te are homogeneous at submicro resolution. Figures 2d and 2e show the elemental distribution of Sn and Te in a small area as indicated by the white rectangle in Figure 2c. Electrical Property. The temperature dependences of electrical resistivity and Seebeck coefficient of CoSb2.75Sn0.25-xTex (x ) 0.125, 0.150, 0.175, 0.200) were shown in Figure 3. Except for the sample with x ) 0.125 showing semiconductor-like behavior, all other samples display semimetal-like character, showing positive temperature dependence on electrical resistivity. The electrical resistivity of CoSb2.75Sn0.25-xTex decreases with increasing Te concentration. Because of the charge compensating effect of Sn, it is considered that only the Te atoms uncompensated by Sn could give out a free electron to contribute to the

Figure 3. Temperature dependence of thermoelectric transport properties for the samples CoSb2.75Sn0.25-xTex (x ) 0.125 0.150, 0.175, 0.200): (a) electrical resistivity, (b) Seebeck coefficient, (c) power factor.

electric transport, and the carrier concentration of CoSb2.75Sn0.05Te0.20 was comparable with CoSb2.85Te0.15. However, the room temperature electrical resistivity of CoSb2.75Sn0.05Te0.20 (10.2 µΩm) is slightly higher than that observed in CoSb2.85Te0.15 (9.5 µΩm).13 This may be caused by the fact that more lattice distortions in CoSb2.75Sn0.05Te0.20 than that in CoSb2.85Te0.15, which generates stronger scattering against the electron, and hence resulting in low mobility. Seebeck coefficient measurements conducted up to 450 °C revealed that all the samples were n-type in the whole temperature range (see Figure 2b). Regardless of composition, the absolute value of Seebeck coefficient (/R/) for CoSb2.75Sn0.25-xTex increases almost linearly with temperature. The highest /R/ near room temperature is ∼182 µV K-1 for the sample with x ) 0.125, which increases almost linearly to ∼221 µV K-1 at 400 °C. As T > 400 °C, /R/ tends to a maximum and then decreases, indicating the onset of intrinsic conduction. With increasing Te content, the extrinsic electron concentration increases while the intrinsic hole decreases originating from the cross-gap excitation.13 As a result, the peak of /R/ shifts toward high temperature with increasing Te content. The temperature-dependent power factor curve shows that the sample with more Te has higher power factor, as shown in Figure 2c. The sample CoSb2.75Sn0.05Te0.20 shows the highest power factor of 3300 µW m-1 K-2 at 450 °C. We also measured the thermoelectric properties of this sample up to ∼550 °C, which showed a higher power factor of 3350 µW m-1 K-2 at 546 °C. This value is slightly lower than the value of 3540 µW m-1 K-2 at 547 °C obtained in CoSb2.85Te0.15.13 By solving Boltzmann transport equation of electron in a parabolic single-band approximate and a nondegenerate

Thermoelectric Performance of CoSb3-xTex

Chem. Mater., Vol. 20, No. 24, 2008 7529

Figure 4. Seebeck coefficient (a) and weighted mobility (b) at room temperature as a function of natural logarithm electrical conductivity for the present samples CoSb2.75Sn0.25-xTex (x ) 0.125 0.150, 0.175, 0.200) and the previous samples CoSb3-xTex (y ) 0, 0.05, 0.10, 0.15, 0.20) from ref 13.

approximate, Seebeck coefficient and electrical conductivity of n-type semiconductors can be expressed as,16 R)-

(

σ ) 2e

2πmokBTo 2

h

kB 5 s+ -ξ e 2

[( ) ]

(1)

T To

(2)

)() 3/2

3/2

(m∗/mo)3/2µ exp(ξ)

where h, ξ, s, To, and mo are the Plank constant, reduced Fermi energy, scattering factor, room temperature and free electron mass, respectively. We can then express the Seebeck coefficient as a function of natural logarithm electrical conductivity R)-

[

()

]

kB 3 T A + ln + ln U - ln σ e 2 To

(3)

where A is scattering factor related parameter which is equal to 17.71 + s, U is weighted mobility which is defined as (m*/mo)3/2µ. The detailed derivation and validity discussion of eq 3 are given out in the Supporting Information. According to eq 3, for a given system with different Fermi energy, the value ∂R/∂ln(σ) should be a classic value kB/e (∼86.2 µV K-1). Figure 4a plots the room temperature Seebeck coefficient of present samples CoSb2.75Sn0.25-xTex and previously reported samples CoSb3-xTex13 as a function of natural logarithm electrical conductivity. Taking the slope of scattering data, we arrive at ∂R/∂ln(σ) ≈ 105 µV K-1 for the CoSb3-xTex samples, and ∂R/∂ln(σ) ≈ 51.5 µV K-1 for the CoSb2.75Sn0.25-xTex samples. Both values significantly deviated from the classic result of ∂R/∂ln(σ)∼kB/e, where kB is the Boltzmann constant and e is the charge of electron. This fact indicates that the carrier effective mass m* or carrier mobility µ is considerably affected by doping. By using eq 3, we estimate the transport parameter U from the measured Seebeck coefficient and electrical resistivity, as shown in (16) Nolas, G. S.; Sharp J.; Goldsmid, H. J. Thermoelectrics: Basic Principles and New Materials DeVelopments; Springer: New York, 2001.

Figure 5. Temperature dependence of (a) total thermal conductivity (filled symbols) and lattice thermal conductivity (unfilled symbols) and temperature dependence of (b) figure of merit for the sample CoSb2.75Sn0.05Te0.20. The date of pure CoSb3 from ref 12 and CoSb2.85Te0.15 from ref 13 are also given for comparison.

Figure 4b. Here, acoustic phonon scattering is taken as the main carrier scattering mechanism, which has been extensively accepted in CoSb3 based materials.7 At first glance, the U values of CoSb4-xTex serial samples are higher than that of CoSb2.75Sn0.25-xTex serial samples. The highest U value is 202 cm2V-1s-1 in CoSb2.95Te0.05, which is comparable to m* ≈ 2.5m0 and µ ≈ 50 cm2 V-1 s-1.7 With increasing ln σ or increasing Te content, the U decreases for the CoSb3-xTex compositions while increases for the CoSb2.75Sn0.25-xTex samples. The U decreases in CoSb3-xTex because of the increased lattice disturbance by Te doping. However, in the CoSb2.75Sn0.25-xTex, the total doping content is the same (i.e., (Te + Sn)/Sb ) 0.25/2.75), U shows a significant dependence on the ratio of Sn/Te. This fact indicates that the point defect SnSb scatters more strongly the transport of electrons than the point defects TeSb. Here, the symbols SnSb and TeSb mean that atoms Sn and Te occupy the Sb site in CoSb3 lattice, respectively. Sn has less valence electron than Sb, whereas Te has more electron than Sb. As a result, the point defect SnSb in CoSb3 lattice is positive and easily absorbs the free electron and hence produces strong scattering. Thermal Conductivity and ZT Value. Figure 5 displays the thermal conductivity and thermoelectric figure of merit for CoSb2.75Sn0.05Te0.20 as a function of temperature. The data of two fine-grained samples CoSb312 and CoSb2.85Te0.1513 fabricated by MA-SPS method in our previous studies were also presented for comparison. It is noted that both samples CoSb3 and CoSb2.85Te0.15 have fine grains (∼160 nm) and hence lower thermal conductivity than the bulk materials with similar composition but large grain size (several micrometers). The thermal conductivity of CoSb2.75Sn0.05Te0.2 is significantly lower than that of CoSb2.85Te0.15 in the whole measured temperature range, as shown in Figure 5a. At room temperature, the thermal conductivity of CoSb2.75Sn0.05Te0.20 is 2.04 Wm-1K-1, which is much lower than that of CoSb2.85Te0.15 (2.66 W m-1 K-1). It is well-known that the bipolar diffusive thermal conductivity (λbip) is significant for

7530

Chem. Mater., Vol. 20, No. 24, 2008

the narrowing semiconductor, especially at high temperature or low doping level. According to our previous study, 13 the λbip/λtot is estimated to be only 2.5% for the CoSb2.85Te0.15 because of the high extrinsic carrier. By considering electric charge compensation of Te by Sn, the donor doping level of CoSb2.75Sn0.05Te0.20 is comparable to that of CoSb2.85Te0.15. Therefore, we estimated λlat by subtracting λcar from λtot, as indicated by unfilled symbols in Figure 5a. Here, λcar can be estimated by Wiedemann-Franz’s law, kcar ) LT/F, where L is calculated to be 1.7 × 10-8 V2 K-2. Because the difference in λcar between CoSb2.75Sn0.05Te0.20 (0.445 W m-1 K-1 at 50 °C) and CoSb2.85Te0.15 (0.518 W m-1 K-1 at 50 °C) is quite small, the reduction in thermal conductivity is mainly contributed to by the decreased λlat because of the enhancement of point defect scattering. The present sample CoSb2.75Sn0.05Te0.20 has similar grain size and density with previous sample CoSb2.85Te0.15, but its lattice thermal conductivity is significantly reduced. The thermal conductivity of CoSb3-based material can be considerably decreased by including mass fluctuation through doping other atoms,7,8 by reducing grain size due to the grain boundary scattering,10,12 or by combining both scattering effects.7,13 The low thermal conductivity of present CoSb2.75Sn0.05Te0.20 is a result of combining the grain boundary scattering and the point defect scattering. Since the electric charge compensation of Sn atoms, the solubility limit of Te in Co(Sb, Sn)3-xTex is shifted from x ) 0.15 to x ) 0.20. The amount of mass fluctuation in CoSb2.75Sn0.05Te0.20 is higher than that in CoSb2.85Te0.15. According to the measured electrical resistivity, Seebeck coefficient, and thermal conductivity, the ZT values were calculated and plotted as a function of temperature in Figure 5b. The ZT values of CoSb2.75Sn0.05Te0.20 are higher than that of CoSb2.85Te0.15 in the whole measured temperature range (50-550 °C) because of significantly decreased lattice thermal conductivity. The maximum ZT value of CoSb2.75Sn0.05Te0.20 is 1.1 at ∼550 °C, which is about 10fold higher than the maximum of pure CoSb3. This value is also comparable to many state-of-the-art filled CoSb3 compounds.4-6 Phonon Scattering Mechanism. As compared with CoSb2.85Te0.15, the enhanced ZT in CoSb2.75Sn0.05Te0.20 is mainly due to the significant decrease in lattice thermal conductivity while slight decrease in electrical conductivity. From the viewpoint of point defect scattering, it sounds reasonable because the sample CoSb2.75Sn0.05Te0.20 have much more mass fluctuations than the sample CoSb2.85Te0.15. We also try to explore the deeper reasons from the viewpoints of the change in vibration mode and the difference in nanostructure nature. Figure 6 shows the Raman shift of CoSb2.86M0.02Te0.12 (M ) Sb, Si, Ge, Sn, Pb). According to the theoretical calculation,17 there are eight Raman sensitive phonon vibration modes in CoSb3, identified as 2Ag + 2Eg + 4Fg, with six appeared in the wavenumber ranges from 100 to 225 cm-1. Five of six modes were observed in our Raman scattering measurements. Since the theoretical Fg4 mode (178 cm-1) and Ag2 mode (179 cm-1) are so close that we were unable to distinguish them. As M ) Sb, i.e., for the CoSb2.88Te0.12 sample, the peaks at 134.4, 150.2, (17) Feldman, J. L.; Singh, D. J. Phys. ReV. B 1996, 53, 6273.

Liu et al.

Figure 6. Raman shift of the samples CoSb2.86M0.02Te0.12 (M ) Sb, Si, Ge, Sn, Pb). The data of theoretically calculated modes for CoSb3 from ref 17 are also presented.

176.5, and 183.5 cm-1 were observed, which are in good agreement with the theoretical predication. The highestfrequency vibration mode shows the highest Raman response intensity, which is inconsistent with Nolas et al.’s observation.18 We believe that this mode should be the Eg2 vibration mode, which is related with the out-of-phase motion between two Sb4 rings. The samples CoSb2.86Si0.02Te0.12 and CoSb2.86Pb0.02Te0.12 display similar Raman response patterns to CoSb2.88Te0.12, whereas the samples CoSb2.86Ge0.02Te0.12 and CoSb2.86Sn0.02Te0.12 show significant changes. The most notable change is the shift of peaks toward the lowfrequency. Koyanagi et al.19 reported that Sb in CoSb3 can be partially substituted by Ge and Sn but not by Pb. Therefore, we deduced that Si and Pb did not enter into the Sb site because of the large atomic size difference, while Ge and Sn got into the Sb site and generated considerable disturbance of vibration mode. In the theoretical Raman vibration modes of CoSb3, the wavenumber distances of Fg3, Fg4, and Ag2 from Eg2 are 25, 4, and 3 cm-1, respectively. Surprisingly, a “new” peak at ∼167.3 cm-1 was observed in both CoSb2.86Ge0.02Te0.12 and CoSb2.86Sn0.02Te0.12, which have ∼11 cm-1 from the Eg2 vibration model (∼178.3 cm-1). This possibly resulted from the shift by one of the Fg3, Fg4, and Ag2 modes, or the splitting of one degenerate vibration mode because of the broken symmetry of Sb4 rings as Sn, Te, or both Sn and Te enter into the Sb4 rings. The effect of point defect scattering on the thermal conductivity can be understood by the significant changes in the vibration modes. We also noted that the CoSb3-y-xMyTex (M ) Ge, Sn) family can be considered as an alloy between binary compound CoSb3 and ternary compound that have a general formula AM1.5Te1.5.20-22 The quaternary compound CoSb3-y-xMyTex is very analogous to the AgPbmXTem+2 (X ) Sb, Bi) system, which is also thought of as an alloy of PbTe and AgXTe2.23 One of the predominant features in AgPbmSbTem+2 systems is the presence of quantum “nano(18) Nolas, G. S.; Kendziora, C. A.; Takizawa, H. J. Appl. Phys. 2003, 94, 7440. (19) Koyanagi, T.; Tsubouchi, T.; Ohtani, M.; Kishimoto, K.; Anno H.; Matsubara. K. Proceedings of the 15th International Conference on Theromoelectrics, Pasadena, CA, March 26-29, 1996; Institute of Electrical and Electronics Engineers: Piscataway, NJ, 1996; p 107. (20) Nolas, G. S.; Yang, J.; Ertenberg, R. W. Phys. ReV. B 2003, 68, 193206. (21) Bos, J. W. G.; Cava, R. J. Solid State Commun. 2007, 141, 38. (22) Vaqueiro, P.; Sobany, G. G.; Stindl, M. J. Solid State Chem. 2008, 181, 768.

Thermoelectric Performance of CoSb3-xTex

Chem. Mater., Vol. 20, No. 24, 2008 7531

Therefore, these “nanodots” would scatter phonon transport and were responsible for the low thermal conductivity of CoSb2.75Sn0.05Te0.20 besides the point defect scattering and the grain boundary scattering. Conclusions

Figure 7. Nanostructure in the single grain of the sample CoSb2.75Sn0.05Te0.20: (a) a panorama of a grain, (b) a possible “nanodot”.

dots” through a melting and slow cooling method.24,25 Wang et al.26 also observed nanosized inhomogenous region in their bulk material fabricated MA-SPS method. The fabrication process of CoSb2.75Sn0.05Te0.20 investigated is similar to that used by Wang et al.. We also expected that a similar microstructure should be observed. Figure 7 shows high resolution TEM images of the CoSb2.75Sn0.05Te0.20 sample. A few nanoscopic black regions embedded in the matrix, as shown in Figure 7a. As seen from Figure 7b, the highresolution TEM image confirmed the character of “nanodot” observed in the AgPb18XTe20 (X ) Sb, Bi) system.24-27 (23) Kanatzidis, M. G. The Role of Solid-State Chemistry in the Discovery of New Thermoelectric Materials. In Recent Trends in Thermoelectric Materials Research I; Tritt, T. M., Ed.; Semiconductors and Semimetals; Academic Press: San Diego, 2001; Vol. 69; p 51. (24) Hsu, K. F.; Loo, S.; Guo, F.; Chen, W.; Dyck, J. S.; Uher, C.; Hogan, T.; Polychroniadis, E. K.; Kanatzidis, M. G. Science 2004, 303, 818. (25) Han, M. K.; Hoang, K.; Kong, H.; Pcionek, R.; Uher, C.; Paraskevopoulos, K. M.; Mahanti, S. D.; Kanatzidis, M. G. Chem. Mater. 2008, 20, 3512. (26) Wang, H.; Li, J. F.; Nan, C. W.; Zhou, M.; Liu, W. S.; Zhang, B. P.; Kita, T. Appl. Phys. Lett. 2006, 88, 092104. (27) Zhou, M.; Li, J. F.; Kita, T. J. Am. Chem. Soc. 2008, 130, 4527.

The solubility limit of Te in CoSb3-xTex is increased by additional substitution of IVB-group elements as electriccharge compensation. It was found that the thermal conductivity (electrical resistivity) of CoSb2.75Sn0.05Te0.20 is much lower (slight higher) than that of CoSb2.85Te0.15. An enhanced ZT value as high as 1.1 at ∼550 °C was therefore obtained in CoSb2.75Sn0.05Te0.20. The Raman scattering measurement of CoSb2.86M0.02Te0.12 (M ) Si, Ge, Sn, Pb) indicates that Ge and Sn enter into the Sb site of CoSb3 and generate significant changes in the vibration modes, whereas Si and Pb are not likely to get into the Sb site of CoSb3. This notable change in atomic vibration mode can be used to explain the effect of point defect scattering against the thermal conductivity. A few “nanodots” similar to that observed in Ag1-xPbmSbTem+2 exist, which are also responsible for the low thermal conductivity. Acknowledgment. This work was supported by National Basic Research Program of China (Grant Nos. 2007CB607504 and 2007CB607505) and Natural Science Foundation of Beijing (Grant 2072009). Supporting Information Available: Calculated heat capacity of pure CoSb3, the derivation detail of weight mobility formula, the validity and possible error of weight mobility formula (PDF). This material is available free of charge via the Internet at http://pubs.acs.org. CM802367F